Skip to Main content Skip to Navigation
New interface
Conference papers

Résolution du shape-from-shading par apprentissage

Abstract : In this paper, we try to solve the shape-from-shading problem using machine learning techniques. A first object whose shape is known is imaged under a known lighting. Its photograph is cut into patches Ia of size 3 × 3, associated with normals N. We process a principal components analysis (PCA) on the data, in order to project the patches Ia in a space of lower dimension n. Then, we analyse the photograph of a second object of unknown shape, taken under the same lighting, which we cut into patches It of the same size 3 × 3. For each patch It, we search its closest patch Ia and assign to It the normal N associated to Ia. This method is tested on synthetic, as well as on real images. We show its advantages, but also its drawbacks, particularly faced to the well-known problem of the concave/convex ambiguity. Nevertheless, this work opens some perspectives for the resolution of the photometric stereo problem.
Document type :
Conference papers
Complete list of metadata

Cited literature [18 references]  Display  Hide  Download
Contributor : Peter Sturm Connect in order to contact the contributor
Submitted on : Friday, May 27, 2011 - 10:10:03 AM
Last modification on : Friday, August 5, 2022 - 2:56:21 PM
Long-term archiving on: : Sunday, August 28, 2011 - 2:21:32 AM


Files produced by the author(s)


  • HAL Id : inria-00596354, version 1


Jean-Denis Durou, Vincent Charvillat, Maxime Daramy, Pierre Gurdjos. Résolution du shape-from-shading par apprentissage. ORASIS - Congrès des jeunes chercheurs en vision par ordinateur, INRIA Grenoble Rhône-Alpes, Jun 2011, Praz-sur-Arly, France. ⟨inria-00596354⟩



Record views


Files downloads